product-health-agent

# Product Health Agent — Managed Observability for Production Apps A Claude Managed Agent that answers two questions about production apps every day: 1. **What parts of the app are users engaging with most?** 2. **What's not functional well operationally?** The agent has: - Custom tools for Cloud Logging, Cloud Monitoring, Firestore aggregation reads - Read access to the app's source repos (mounted at session start) - The standard agent toolset (bash, read, write, grep, glob, web_search) - **Zero write access to production** — only the workspace insights collection ## Status **v1 — built 2026-04-11.** Supports Glam and SpinDineSwift on Firebase. Ready to run daily digests and ad-hoc deep dives. ## Architecture ``` ┌──────────────────────────────────────────────────────────────┐ │ Host (cron or human invocation) │ │ │ │ digest_runner.py / orchestrator.py │ │ ├── google-cloud-firestore creds (ADC) │ │ ├── google-cloud-logging creds (ADC) │ │ ├── google-cloud-monitoring creds (ADC) │ │ └── tools.py dispatch() — host-side tool execution │ │ │ │ Sessions are created per-run. Agent is persistent. │ └───────────────────────┬──────────────────────────────────────┘ │ streams events, dispatches c...